RECIFE: a MCDSS for Railway Capacity Xavier GANDIBLEUX (1), Pierre RITEAU (1), and Xavier DELORME (2) (1) LINA - Laboratoire d Informatique de Nantes Atlantique Universite de Nantes 2 rue de la Houssiniere BP92208, F-44322 Nantes cedex 03 FRANCE Xavier.Gandibleux@univ-nantes.fr, Pierre.Riteau@etu.univ-nantes.fr (2) Centre Génie Industriel et Informatique École des Mines de Saint-Etienne 158 cours Fauriel, F-42023 Saint-Etienne cedex 2 FRANCE Delorme@emse.fr MCDM 08 19th International Conference on Multiple Criteria Decision Making January 07-13, 2008 Auckland, New-Zealand
RECIFE: a MCDSS for railway capacity (#2) Objectives of RECIFE project Research project with INRETS and SNCF: Models and algorithms to evaluate railway infrastructure capacity Tools integrated in a decision support software Application on Pierrefitte-Gonesse node (junction) Lille-Flandres (station)
RECIFE: a MCDSS for railway capacity (#3) 1. Railway Infrastructure Operation Planning 2. Information System 3. Screenshots 4. Conclusion
RECIFE: a MCDSS for railway capacity (#4) 1.1 Problematic: a railway planning problem (RPP) Planning the construction or reconstruction of infrastructures
RECIFE: a MCDSS for railway capacity (#5) 1.1 Problematic: a railway planning problem (RPP) Planning the construction or reconstruction of infrastructures
RECIFE: a MCDSS for railway capacity (#6) 1.1 Problematic: a railway planning problem (RPP) Planning the construction or reconstruction of infrastructures
RECIFE: a MCDSS for railway capacity (#7) 1.1 Problematic: a railway planning problem (RPP) Planning the construction or reconstruction of infrastructures Capacity of one component / junctions of a rail system + 8 < : Safety rules Rolling stock technical characteristics Service quality How many trains can be routed through the junction within a time interval? What is the best solution to route these trains?
RECIFE: a MCDSS for railway capacity (#8) 1.1 Problematic: a railway planning problem (RPP) Planning the construction or reconstruction of infrastructures Capacity of one component / junctions of a rail system Junction Pierrefitte-Gonesse, north of Paris
RECIFE: a MCDSS for railway capacity (#9) 1.2 Decision process Helping the decision-maker (expert in railway management) to answer to the feasibility and/or saturation problem plus the stability problem (ability to absorb delays) Decision process structured lexicographically by two criteria: 1st criterion: max the number of train 2nd criterion: max the stability among the equivalent timetables (minimize the sum of delays)
RECIFE: a MCDSS for railway capacity (#10) 1. Railway Infrastructure Operation Planning 2. Information System 3. Screenshots 4. Conclusion
RECIFE: a MCDSS for railway capacity (#11) 2.1 Organisation of the information system
RECIFE: a MCDSS for railway capacity (#12) 2.2 Input data: one situation Kind of traffic, time-windows in the day, density, etc. Data (infrastructure, service, rolling stock, safety rules) All possible routes are given All possible arrival-date are given Resource comsumed: One list of trains
RECIFE: a MCDSS for railway capacity (#13) 2.2 Input data: one situation
RECIFE: a MCDSS for railway capacity (#14) 2.3 Handling the first criterion: optimization stage Given a finite set I = {1,...,n} of items {T j },j J = {1,...,m}, a collection of m subsets of I a packing is a subset P I such that T j P 1, j J which 2 Max z(x) = X 3 c i x i i I X t i,j x i 1, j J i I (SPP) 6 x i {0,1}, i I 7 4 5 t i,j {0,1}, i I, j J Set Packing Prb (SPP): strongly NP-Hard (Garey and Johnson 1979) Solvers: exact, Cplex; metaheuristics, GRASP; ACO
RECIFE: a MCDSS for railway capacity (#15) 2.3 Handling the first criterion: optimization stage t in Road 1 Road 2 TGV Road 1 Road 2 Road 3 t in Road 3 t in Road 1 Road 2 TGV Road 1 Road 2 Road 3 t in Road 3
RECIFE: a MCDSS for railway capacity (#16) 2.3 Handling the first criterion: optimization stage
RECIFE: a MCDSS for railway capacity (#17) 2.4 Data in output A solution: a list L of timetables equivalent timetables: same number of trains different timetables: infrastructure used, trains selected (saturation), etc.
RECIFE: a MCDSS for railway capacity (#18) 2.5 Handling the second criterion: simulation stage The simulation and analysis modules: help the decision-maker to evaluate the stability of the generated timetables to determine the critical items Principle (1/2): delay propagation Two types of delay primary delay caused by a disruption secondary delay due to interactions between trains Impact of a primary delay secondary delays generated directly or indirectly only short primary delay considered Processing the conflicts arrival-date of other trains delayed routes and schedules maintained (no re-optimization)
RECIFE: a MCDSS for railway capacity (#19) 2.5 Handling the second criterion: simulation stage Principle (2/2): delay propagation Measure the effect Domino effect: sum of secondary delays of a primary delay Series of shortest path computation Illustration: didactic example on Pierrefitte-Gonesse node: 5 trains routed, 12 different timetables generated Stability evaluation: One graph of potential direct conflicts for each timetable: 2 primary delay values (180s & 300s). For the primary delay of 180s:
RECIFE: a MCDSS for railway capacity (#20) 2.5 Handling the second criterion: simulation stage
RECIFE: a MCDSS for railway capacity (#21) 2.5 Handling the second criterion: simulation stage
RECIFE: a MCDSS for railway capacity (#22) 2.5 Handling the second criterion: simulation stage Representation in the outcome space 12 timetables 3 potentially efficient solutions
RECIFE: a MCDSS for railway capacity (#23) 2.5 Handling the second criterion: simulation stage Principle: the DM simulates the effect of delays (1/2) to assess the stability: primary delay one objective set of objectives dynamically defined (what-if) analyse of efficient timetables visual analyse - global comparizon (performances in the outcome space) - local comparison of k-efficient sols (perfs on criteria) quantitative analyse - pairwize comparison (solutions) - statistics of resources used (solution) - statistics on delay propagated [critical train] (solution)
RECIFE: a MCDSS for railway capacity (#24) 2.5 Handling the second criterion: simulation stage Principle: the DM simulates the effect of delays (2/2) to catch the uncertainty/incompleteness of the information (data, model) handled: in the objective space, analyse the solutions of rank > 1 to validate a solution in its technical environment a solution is viewed inside the usual graphics handled by the DM (space-time graphic, gantt chart, simulation of traffic on the infrastructure) Data in output: one realistic timetable, which maximizes the number of trains using the infrastructure, for the given scenario of traffic, with a good stability faced to possible delays
RECIFE: a MCDSS for railway capacity (#25) 1. Railway Infrastructure Operation Planning 2. Information System 3. Screenshots 4. Conclusion
RECIFE: a MCDSS for railway capacity (#26) 3.1 Screenshots: focussed on the MCDM aspects
RECIFE: a MCDSS for railway capacity (#27) 3.1 Screenshots: focussed on the MCDM aspects
RECIFE: a MCDSS for railway capacity (#28) 3.2 Screenshots: focussed on a solution
RECIFE: a MCDSS for railway capacity (#29) 3.2 Screenshots: focussed on a solution
RECIFE: a MCDSS for railway capacity (#30) 3.2 Screenshots: focussed on the core of the trade
RECIFE: a MCDSS for railway capacity (#31) 1. Railway Infrastructure Operation Planning 2. Information System 3. Screenshots 4. Conclusion
RECIFE: a MCDSS for railway capacity (#32) 4. Conclusion An optimization model for feasibility and/or saturation set packing problem ant colony optimization based algorithm list of equivalent (but different) railway timetables A multiobjective model for stability evaluation delay propagation method shortest path computation multi-criteria analysis Both integrated in an information system for railway capacity evaluation of junction or station Future research works: multiobjective optimization: search for compromises between capacity use and stability preferences on the traffic integrated in the timetables
RECIFE: a MCDSS for Railway Capacity Xavier GANDIBLEUX, Pierre RITEAU, and Xavier DELORME X. Gandibleux, X. Delorme, and V. T Kindt. An ant colony algorithm for the set packing problem. In M. Dorigo, M. Birattari, Ch. Blum, L. Gambardella, Fr. Mondada, and Th. Stutzle, eds, Ant Colony Optimization and Swarm Intelligence, LNCS 3172, pp. 49 60. Springer, 2004. X. Delorme, X. Gandibleux, and J. Rodriguez. Stability evaluation of a railway timetable at station level. European Journal of Operational Research, 2007. (Available online, doi:10.1016/j.ejor.2007.06.062). J. Rodriguez, X. Delorme, X. Gandibleux, Gr. Marlière, R. Bartusiak, F. Degoutin, and S. Sobieraj. RECIFE: models and tools for analyzing rail capacity. Recherche Transports Sécurité, 95:19 36, 2007. (doi:3166/rts.95.19-36).